Training method, device and equipment of power distribution network bearing capacity evaluation model

Abstract

The invention relates to a training method, device and equipment for a power distribution network bearing capacity evaluation model. The method comprises the following steps: constructing an initial power distribution network bearing capacity evaluation model, wherein the initial power distribution network bearing capacity evaluation model comprises a preorder processing network and a back propagation neural network; the preorder processing network is constructed based on a genetic algorithm or a sparrow search algorithm; acquiring data of the power distribution network under each bearing capacity evaluation index after the plurality of charging piles are merged into the power distribution network; and training the initial power distribution network bearing capacity evaluation model based on the data of the power distribution network under each bearing capacity evaluation index, and obtaining the power distribution network bearing capacity evaluation model after the training is finished. By adopting the method, the accuracy of evaluating the bearing capacity of the power distribution network can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of distribution network carrying capacity evaluation, and particularly to a training method, device and equipment for a distribution network carrying capacity evaluation model. Background Art

[0002] With the rapid growth in the number of electric vehicles, the charging demand is increasing day by day. Correspondingly, the scale of charging piles is also increasing. The access of a large number of charging piles will have a significant impact on the distribution network, such as increasing the peak-valley difference of the distribution network load, voltage deviation and line overload.

[0003] Therefore, with the growth of charging demand, fully considering the impact of charging pile grid connection on the distribution network to ensure the safe and stable operation of the power grid has become an urgent issue. In the process of determining the impact of charging pile grid connection on the distribution network in traditional solutions, there is a problem of inaccurate evaluation of the carrying capacity of the distribution network.

[0004] How to improve the accuracy of the evaluation of the carrying capacity of the distribution network remains to be solved urgently. Summary of the Invention

[0005] Based on this, in view of the above technical problems, it is necessary to provide a training method, device and equipment for a distribution network carrying capacity evaluation model that can improve the accuracy of the evaluation of the carrying capacity of the distribution network.

[0006] In a first aspect, the present application provides a training method for a distribution network carrying capacity evaluation model, the method comprising:

[0007] Construct an initial distribution network carrying capacity evaluation model, the initial distribution network carrying capacity evaluation model includes a pre-processing network and a backpropagation neural network, and the output of the pre-processing network is used as the input of the backpropagation neural network; the pre-processing network is constructed based on a genetic algorithm or a sparrow search algorithm;

[0008] Obtain data of the distribution network under each carrying capacity evaluation index after multiple charging piles are incorporated into the distribution network; wherein, the carrying capacity evaluation index includes at least one of the N-1 passing rate of medium-voltage lines, average power outage frequency, average outage duration, evaluation power supply reliability rate, voltage qualification rate, voltage deviation, total harmonic distortion rate, line loss rate, line maximum load rate, line annual average load rate, distribution transformer annual average load rate, net load maximum allowable volatility, net load volatility or load transfer ratio under different carrying capacity levels;

[0009] Train the initial distribution network carrying capacity evaluation model based on the data of the distribution network under each carrying capacity evaluation index, and obtain the distribution network carrying capacity evaluation model after the training is completed.

[0010] In one embodiment, when the preprocessing network is constructed based on a genetic algorithm, the fitness function of the preprocessing network is used to take the reciprocal of the prediction error of the backpropagation neural network as the individual fitness value;

[0011] During the process of training the initial distribution network carrying capacity evaluation model, the processing process of the preprocessing network includes:

[0012] According to the coding range of [-1, 1], the connection weights between the input layer and the hidden layer, the critical activation thresholds of the neurons in the hidden layer, the connection weights between the hidden layer and the output layer, and the critical activation thresholds of the neurons in the output layer in the backpropagation neural network are respectively encoded as chromosomes, and an initial population is formed based on the encoded chromosomes; where one individual in the population corresponds to one chromosome;

[0013] Determine the individual fitness of each individual in the initial population based on the fitness function, and iteratively optimize the initial population based on the individual fitness of each individual until the termination condition is reached, and then output the optimal individual in the optimized population;

[0014] Correspondingly, the processing process of the backpropagation neural network includes:

[0015] Decode the optimal individual, and update the network structure parameters based on the optimal connection weights between the input layer and the hidden layer, the optimal critical activation thresholds of the neurons in the hidden layer, the optimal connection weights between the hidden layer and the output layer, and the optimal critical activation thresholds of the neurons in the output layer obtained by decoding;

[0016] In one embodiment, when the preprocessing network is constructed based on a sparrow search algorithm, the fitness function of the preprocessing network is used to take the reciprocal of the prediction error of the backpropagation neural network as the individual fitness value;

[0017] During the process of training the initial distribution network carrying capacity evaluation model, the processing process of the preprocessing network includes:

[0018] Take the sum of the number of nodes in the input layer, the hidden layer, and the output layer in the backpropagation neural network as the number of sparrow individuals in the initial population, and construct an initial population;

[0019] Encode the population positions of the sparrow individuals in the population as vectors; the vectors correspond to the connection weights between the input layer and the hidden layer, the critical activation thresholds of the neurons in the hidden layer, the connection weights between the hidden layer and the output layer, and the critical activation thresholds of the neurons in the output layer in the backpropagation neural network;

[0020] Determine the individual fitness of sparrow individuals in the initial population based on the fitness function, and iteratively optimize the initial population based on the individual fitness of each individual until the termination condition is reached, and then output the optimal vector of the sparrow individuals in the optimized population;

[0021] Correspondingly, the processing process of the backpropagation neural network includes:

[0022] Decode the optimal vector, and update the network structure parameters based on the optimal connection weights between the input layer and the hidden layer, the optimal critical activation thresholds of the neurons in the hidden layer, the optimal connection weights between the hidden layer and the output layer, and the optimal critical activation thresholds of the neurons in the output layer obtained by decoding.

[0023] In one embodiment, the backpropagation neural network includes an input layer, an output layer, and one hidden layer; wherein, the hidden layer includes 3 neurons, the activation function of the hidden layer is the hyperbolic tangent function, and the activation function of the output layer is a linear function.

[0024] In one embodiment, training the initial distribution network carrying capacity evaluation model based on the data of the distribution network under each carrying capacity evaluation index includes:

[0025] Merge the data of the distribution network under each carrying capacity evaluation index into a data set, and divide the data set into a training set and a test set;

[0026] Use the data in the test set as the initial input data, normalize the initial input data to the data within the range composed of 0 to 1, and add random noise to the normalized data to obtain the input data;

[0027] Train the initial distribution network carrying capacity evaluation model according to the input data and the test set.

[0028] In a second aspect, the present application provides a distribution network carrying capacity evaluation method, and the method includes:

[0029] Obtain the distribution network carrying capacity evaluation model trained according to the training method of the distribution network carrying capacity evaluation model described in the first aspect;

[0030] Obtain the data of the distribution network under each carrying capacity evaluation index after multiple charging piles are incorporated into the distribution network; wherein, the carrying capacity evaluation index includes at least one of the N-1 passing rate of medium-voltage lines, average power outage frequency, average outage duration, evaluation power supply reliability rate, voltage qualification rate, voltage deviation, total harmonic distortion rate, line loss rate, line maximum load rate, line annual average load rate, distribution transformer annual average load rate, net load maximum allowable volatility, net load volatility, or load transfer ratio under different carrying capacity levels;

[0031] Input the data of the distribution network under each carrying capacity evaluation index into the distribution network carrying capacity evaluation model to obtain the carrying capacity level output by the distribution network carrying capacity evaluation model; wherein, the carrying capacity level is positively correlated with the carrying capacity of the distribution network.

[0032] In a third aspect, the present application provides a training device for a distribution network carrying capacity evaluation model, and the device includes:

[0033] A model construction module, configured to construct an initial distribution network carrying capacity evaluation model, where the initial distribution network carrying capacity evaluation model includes a preprocessing network and a backpropagation neural network, and the output of the preprocessing network is used as the input of the backpropagation neural network; the preprocessing network is constructed based on a genetic algorithm or a sparrow search algorithm;

[0034] An acquisition module, configured to acquire the data of the distribution network under each carrying capacity evaluation index after multiple charging piles are incorporated into the distribution network; wherein, the carrying capacity evaluation index includes at least one of the passing rate of medium-voltage line N-1, average power outage frequency, average outage duration, evaluation power supply reliability rate, voltage qualification rate, voltage deviation, total harmonic distortion rate, line loss rate, maximum line load rate, annual average line load rate, annual average distribution transformer load rate, maximum allowable volatility of net load, volatility of net load, or load transfer ratio under different carrying capacity levels;

[0035] A training module, configured to train the initial distribution network carrying capacity evaluation model based on the data of the distribution network under each carrying capacity evaluation index, and obtain the distribution network carrying capacity evaluation model after the training is completed.

[0036] In a fourth aspect, the present application provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, it implements the steps of the method described in the first aspect, or implements the steps of the method described in the second aspect.

[0037] In a fifth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the method described in the first aspect, or implements the steps of the method described in the second aspect.

[0038] In a sixth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the steps of the method described in the first aspect, or implements the steps of the method described in the second aspect.

[0039] The training method, device, and equipment of the above-mentioned distribution network carrying capacity evaluation model introduce the genetic algorithm on the basis of the backpropagation neural network to construct an initial distribution network carrying capacity evaluation model, namely the GA-BP model, aiming to overcome the limitation that the backpropagation neural network is prone to falling into local optimal solutions. Through the intervention of the genetic algorithm, the randomness of the weight and threshold assignment of the backpropagation neural network is effectively reduced, and the training efficiency of the model is significantly improved. The backpropagation neural network optimized by the genetic algorithm not only enhances the adaptability and convergence of the original backpropagation neural network but also greatly improves the recognition accuracy of the backpropagation neural network, thus enabling the trained distribution network carrying capacity evaluation model to exhibit more excellent performance characteristics.

[0040] Introduce the sparrow search algorithm on the basis of the backpropagation neural network to construct an initial distribution network carrying capacity evaluation model, namely the SSA-BP model. When optimizing the backpropagation neural network, the sparrow search algorithm optimizes the weights and biases of the backpropagation neural network by simulating the foraging and anti-predation behaviors of the sparrow population. In the algorithm, the sparrow population is divided into producers and foragers. The producers are responsible for global search, and the foragers are responsible for local search. The two work together to balance exploration and exploitation. When a predator appears, the sparrow population will adjust its position through an alarm signal to avoid falling into local optimality. In this way, the sparrow search algorithm dynamically adjusts the parameters of the backpropagation neural network in the solution space, thereby improving the convergence speed and prediction accuracy of the network and effectively avoiding the problem that the backpropagation neural network falls into local optimal solutions.

[0041] Therefore, it solves the problem that there are some limitations in the actual application of the backpropagation neural network. If its initial weights and thresholds are randomly set, it is easy to fall into local optimal solutions, resulting in limited performance of the distribution network carrying capacity evaluation model. In this way, the performance of the distribution network carrying capacity evaluation model is improved, making the results obtained when using the distribution network carrying capacity evaluation model to evaluate the distribution network carrying capacity more objective and accurate. Compared with the inaccurate evaluation of the distribution network carrying capacity in the traditional scheme, the method provided in this embodiment improves the accuracy of the evaluation of the distribution network carrying capacity. Brief Description of the Drawings

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for the description in the embodiments of the present application or related technologies. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0043] Figure 1 It is an application environment diagram of the training method of the distribution network carrying capacity evaluation model in an embodiment;

[0044] Figure 2 It is a schematic flowchart of a training method for a distribution network carrying capacity evaluation model in an embodiment;

[0045] Figure 3 It is a partial schematic flowchart of a training method for a distribution network carrying capacity evaluation model in an embodiment;

[0046] Figure 4 It is a partial schematic flowchart of a training method for a distribution network carrying capacity evaluation model in another embodiment;

[0047] Figure 5 It is a partial schematic flowchart of a training method for a distribution network carrying capacity evaluation model in yet another embodiment;

[0048] Figure 6 It is a schematic flowchart of a distribution network carrying capacity evaluation method in an embodiment;

[0049] Figure 7 It is a structural block diagram of a training device for a distribution network carrying capacity evaluation model in an embodiment;

[0050] Figure 8 It is a structural block diagram of a distribution network carrying capacity evaluation device in an embodiment;

[0051] Figure 9 It is an internal structure diagram of a computer device in an embodiment. Detailed implementation manners

[0052] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0053] The training method for the distribution network carrying capacity evaluation model provided by the embodiments of the present application can be applied to an application environment as shown in Figure 1 In the figure, the terminal 102 communicates with the server 104 through a network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or can be placed in the cloud or other network servers. Among them, the terminal 102 can be, but is not limited to, the master station of the distribution network system, the information collection device of the distribution network system (such as a smart meter, a sensing device, etc.). The server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0054] In an exemplary embodiment, as shown in Figure 2As shown, a training method for a distribution network carrying capacity evaluation model is provided. Taking the application of this method to the server 104 in Figure 1 as an example for illustration, it includes the following steps 202 to step 206. Among them:

[0055] Step 202, construct an initial distribution network carrying capacity evaluation model. The initial distribution network carrying capacity evaluation model includes a pre-processing network and a backpropagation neural network. The output of the pre-processing network is used as the input of the backpropagation neural network. The pre-processing network is constructed based on a genetic algorithm or a sparrow search algorithm.

[0056] Among them, the backpropagation neural network (abbreviated as BPNN) is a multi-layer feedforward neural network architecture trained using the error backpropagation algorithm. Please refer to Figure 3 the structural schematic diagram of the backpropagation neural network shown. The network structure consists of an input layer, one or more hidden layers, and an output layer. Each layer contains a certain number of neuron nodes. BPNN continuously optimizes and adjusts the weights and thresholds in the network through two closely related processes: forward propagation of signals and backpropagation of errors, so as to achieve accurate modeling and learning of complex functions. In the operation mechanism of BPNN, signals first enter the network from the input layer, and after weighted operations, they are transmitted to the hidden layer. Each neuron in the hidden layer performs a weighted sum of the input data, inputs the result into the activation function for non-linear transformation, and then continues to transmit the transformed result to the output layer. The output layer receives all inputs from the hidden layer, performs a weighted sum in the same way, and is processed by the activation function to finally generate the actual output value of the network. When there is a significant deviation between the actual output and the expected output, the backpropagation process of errors will be triggered. In the training stage, since there is an error between the output result generated by training and the actual result, it is necessary to backpropagate this error along the network structure to adjust the connection weights and thresholds between layers. This process aims to gradually reduce the error along the direction of gradient descent. After multiple iterations of learning and training, the network gradually converges and finally determines a set of optimal network parameters.

[0057] The backpropagation neural network is prone to falling into local optimal solutions and is significantly affected by the gradient descent algorithm and the initial values. The random setting of the initial weights and thresholds may increase the risk of falling into local minima and affect the learning efficiency of the model. In addition, the generalization ability of the model highly depends on the representativeness of the training samples. Improper sample selection will greatly reduce the prediction accuracy. Therefore, optimizing for these problems is the key direction for future improvement of the backpropagation neural network.

[0058] Among them, the Genetic Algorithm (GA for short) and the Sparrow Search Algorithm (SSA for short) both belong to bionic optimization algorithms. The genetic algorithm simulates biological genetic evolution and optimizes individuals to find the optimal solution through selection, crossover, and mutation. The sparrow search algorithm simulates the foraging and anti-predation behaviors of sparrows to adjust the search strategy. Compared with traditional optimization algorithms, bionic optimization algorithms have strong global search capabilities in complex solution spaces, are not easily trapped in local optima, and also have good self-adaptability and robustness. They can effectively solve the dilemmas of high computational complexity and easy entrapment in local optima faced by traditional algorithms when dealing with complex problems, providing new ideas for solving complex optimization problems. A preprocessing network is constructed based on the genetic algorithm, and the initially constructed distribution network carrying capacity evaluation model can also be called the GA-BP model. A preprocessing network is constructed based on the sparrow search algorithm, and the initially constructed distribution network carrying capacity evaluation model can also be called the SSA-BP model.

[0059] First, based on the national standard limits of the carrying capacity evaluation indicators and the scores given by experts, the evaluation indicators of the distribution network carrying capacity are classified, and then a preprocessing network is created according to the genetic algorithm and the sparrow search algorithm. The BP neural network is optimized based on the preprocessing network to construct the initially constructed distribution network carrying capacity evaluation model. Further, the feasibility of the initially constructed distribution network carrying capacity evaluation model can be verified based on an example of a certain area with distributed power sources and electric vehicle inverters connected to the grid.

[0060] Step 204: After multiple charging piles are incorporated into the distribution network, obtain the data of the distribution network under each carrying capacity evaluation indicator.

[0061] Among them, the charging pile is used to charge new energy vehicles. Preferably, the charging pile provided in this embodiment is a supercharger, which is a device using high-power charging technology and can supplement a large amount of electric energy for new energy vehicles in a short time.

[0062] Among them, the carrying capacity evaluation indicators include at least one of the passing rate of medium-voltage line N-1 under different carrying capacity levels, average power outage frequency, average outage duration, evaluation of power supply reliability rate, voltage qualification rate, voltage deviation, total harmonic distortion rate, line loss rate, maximum line load rate, annual average line load rate, annual average distribution transformer load rate, maximum allowable volatility of net load, volatility of net load, or load transfer ratio.

[0063] Specifically, first obtain the following fourteen indicators: the passing rate of medium-voltage line N-1, the average power outage frequency, the average outage duration, the evaluated power supply reliability rate, the voltage qualification rate, the voltage deviation, the total harmonic distortion rate, the line loss rate, the maximum line load rate, the annual average line load rate, the annual average distribution transformer load rate, the maximum allowable volatility of net load, the net load volatility, and the load transfer ratio. Then, according to national standards and the actual operation of each indicator, expert assignment is carried out. The bearing capacity evaluation indicators are divided into five levels: I, II, III, IV, and V. Among them, I is the excellent level, II and III are good, IV is qualified, and V is unqualified, decreasing successively.

[0064] The grading of these fourteen indicators under different bearing capacity levels is shown in Table 1.

[0065] Table 1:

[0066] Evaluation Index Ⅰ Ⅱ Ⅲ Ⅳ Ⅴ Pass Rate of Medium-Voltage Line N-1 (%) 80-100 60-80 40-60 20-40 ≤20 Average Power Outage Frequency ≤0.17 0.17-0.19 0.19-0.21 0.21-0.25 ≥0.25 Average Outage Duration ≤0.72 0.72-1.51 1.51-3.01 3.01-3.58 ≥3.58 Evaluated Power Supply Reliability Rate (%) ≥99.971 99.952-99.971 99.931-99.951 99.918-99.931 ≤99.918 Voltage Qualification Rate (%) 99-100 98-99 97-98 96-97 ≤96 Voltage Deviation 0-2 2-3 3-4 4-5 5-7 Total Harmonic Distortion Rate 0-1.5 1.5-2 2-3 3-4 ≥4 Line Loss Rate 0-1.4 1.5-2.9 3-4.4 4.5-4.9 6-7 Maximum Line Load Rate (%) ≥60 46-60 31-45 16-30 0-15 Annual Average Line Load Rate (%) ≥60 46-60 31-45 16-30 0-15 Annual Average Distribution Transformer Load Rate (%) 65-70 60-65 50-60 40-50 0-40 Maximum Allowable Volatility of Net Load (%) 80-100 60-80 40-60 20-40 0-20 Volatility of Net Load (%) ≥40 30-40 20-30 10-20 0-10 Load Transfer Ratio (%) 80-100 60-80 40-60 20-40 0-20

[0067] Step 206: Based on the data of the distribution network under each bearing capacity evaluation indicator, train the initial distribution network bearing capacity evaluation model. After the training is completed, obtain the distribution network bearing capacity evaluation model.

[0068] That is to say, train the initial distribution network bearing capacity evaluation model with the data of the distribution network under all bearing capacity evaluation indicators. The result output by the distribution network bearing capacity evaluation model is the bearing capacity level of the distribution network.

[0069] Optionally, use the data of the distribution network under all bearing capacity evaluation indicators as the training set to train the initial distribution network bearing capacity evaluation model. After reaching the training end condition, obtain the trained distribution network bearing capacity evaluation model. Optionally, divide the data of the distribution network under all bearing capacity evaluation indicators into a training set and a test set. Use the training set to train the initial distribution network bearing capacity evaluation model and use the test set to evaluate each training result. In the case where the training result does not reach the training end condition, repeatedly train the initial distribution network bearing capacity evaluation model until the training end condition is reached, and obtain the trained distribution network bearing capacity evaluation model.

[0070] For example, for five different load-carrying capacity levels (labeled as Ⅰ, Ⅱ, Ⅲ, Ⅳ, and Ⅴ respectively), 15 different evaluation indicators were generated for each load-carrying capacity level. The values under each evaluation indicator could be randomly generated using the randi function within the specified value range. Each evaluation indicator under each load-carrying capacity level contained 50 data points, simulating the sample quantity in the actual dataset. Further, to increase the diversity and authenticity of the data, random noise was added to the values of each evaluation indicator based on the generated original data. The range of the noise was between -10 to 10 and -20 to 20 respectively. Although the noise was also randomly generated, their values were relatively fixed, reflecting the attributes or labels unique to certain load-carrying capacity levels. Finally, the data of all load-carrying capacity levels were combined into a large dataset, which contained 250 samples (50 samples for each load-carrying capacity level) and 15 evaluation indicators. The dataset was randomly divided into a training set (200 samples) and a test set (50 samples) in an 8:2 ratio to ensure the generalization of the distribution network load-carrying capacity evaluation model.

[0071] In the above training method of the distribution network load-carrying capacity evaluation model, a genetic algorithm was introduced based on the backpropagation neural network to construct an initial distribution network load-carrying capacity evaluation model, namely the GA-BP model, aiming to overcome the limitation that the backpropagation neural network is prone to falling into local optimal solutions. Through the intervention of the genetic algorithm, the randomness of the weight and threshold assignment of the backpropagation neural network was effectively reduced, and the training efficiency of the model was significantly improved. The backpropagation neural network optimized by the genetic algorithm not only enhanced the adaptability and convergence of the original backpropagation neural network but also greatly improved the recognition accuracy of the backpropagation neural network, thus enabling the trained distribution network load-carrying capacity evaluation model to exhibit more excellent performance characteristics.

[0072] A sparrow search algorithm was introduced based on the backpropagation neural network to construct an initial distribution network load-carrying capacity evaluation model, namely the SSA-BP model. When optimizing the backpropagation neural network, the sparrow search algorithm optimized the weights and biases of the backpropagation neural network by simulating the foraging and anti-predation behaviors of the sparrow population. In the algorithm, the sparrow population was divided into producers and foragers. The producers were responsible for global search, and the foragers were responsible for local search. The two worked together to balance exploration and exploitation. When a natural enemy appeared, the sparrow group would adjust its position through an alarm signal to avoid falling into local optimality. In this way, the sparrow search algorithm dynamically adjusted the parameters of the backpropagation neural network in the solution space, thereby improving the convergence speed and prediction accuracy of the network and effectively avoiding the problem that the backpropagation neural network falls into local optimal solutions.

[0073] Therefore, the method provided in this embodiment solves some limitations existing in the practical application of the backpropagation neural network. If the initial weights and thresholds are randomly set, it is easy to fall into a local optimal solution, resulting in limited performance of the distribution network carrying capacity evaluation model. In this way, the performance of the distribution network carrying capacity evaluation model is improved, making the results obtained when using the distribution network carrying capacity evaluation model to evaluate the distribution network carrying capacity more objective and accurate. Compared with the inaccurate evaluation of the distribution network carrying capacity by the traditional scheme, the method provided in this embodiment improves the accuracy of the evaluation of the distribution network carrying capacity.

[0074] In some embodiments, the process of training the initial distribution network carrying capacity evaluation model is further described.

[0075] In the case where the preprocessing network is constructed based on the genetic algorithm, the fitness function of the preprocessing network is used to take the reciprocal of the prediction error of the backpropagation neural network as the individual fitness value.

[0076] At the same time, please refer to Figure 3 the flowchart of the genetic algorithm for optimizing the backpropagation neural network shown in. During the process of training the initial distribution network carrying capacity evaluation model, the processing process of the preprocessing network includes:

[0077] Step 1, according to the coding range of [-1, 1], encode the connection weights between the input layer and the hidden layer, the critical activation thresholds of the neurons in the hidden layer, the connection weights between the hidden layer and the output layer, and the critical activation thresholds of the neurons in the output layer in the backpropagation neural network into chromosomes respectively, and form an initial population based on the encoded chromosomes; among them, an individual in the population corresponds to a chromosome.

[0078] The first step is sample preprocessing. The preprocessing network performs normalization processing on the input data to eliminate the dimensional difference and ensure that the distribution of the input data meets the input requirements of the backpropagation neural network. Optionally, the data of the distribution network under all carrying capacity evaluation indicators can be divided into a training set, a validation set and a test set, and the ratio is determined according to the characteristics of the specific data set to ensure the generalization ability of the model.

[0079] The second step is to determine the structure of the backpropagation neural network. Design the structure of the backpropagation neural network as a single-hidden-layer feedforward neural network, and the hidden layer contains 5 neurons. Optionally, the activation function uses the hyperbolic tangent function (tansig), and the output layer uses the linear function (purelin). The number of nodes in the input layer and the output layer is determined by the data feature dimension. The number of nodes in the hidden layer is 5, and the number of nodes in the output layer is 1. The total number of optimization variables can be determined according to the formula determined. In the formula, represents the total number of optimization variables, represents the number of nodes in the input layer, represents the number of hidden layer nodes, represents the number of output layer nodes.

[0080] Step 3: Encode the weights and thresholds of the backpropagation neural network into chromosomes, where each chromosome corresponds to an individual (an individual can also be called an optimization variable). The length of the chromosome is determined by the total number of optimization variables. If the number of input layer nodes of the backpropagation neural network is 3, then the chromosome length is equal to 3×5 + 5 + 5×1 + 1 = 26. The encoding range is restricted to [-1, 1], which can prevent parameter explosion.

[0081] Step 4: Initialize the genetic parameters. Specifically, set the population size to 5, the maximum number of generations to 50, and the optimization variable boundary to [-1, 1]. A small population size reduces the computational cost and is suitable for medium and small datasets. The selection operation is geometric distribution selection (probability 0.09), and elite individuals are retained. The crossover operation is arithmetic crossover (coefficient 2), which enhances the local search ability. The mutation operation is non-uniform mutation (parameters [2, 50, 3]), which balances global and local exploration.

[0082] Step 2: Determine the individual fitness of each individual in the initial population based on the fitness function, and iteratively optimize the initial population based on the individual fitness of each individual until the termination condition is reached, and then output the optimal individual in the optimized population.

[0083] Step 5: The fitness function of the preprocessing network is used to take the reciprocal of the prediction error of the backpropagation neural network as the individual fitness value. That is, Fitness = 1 / (MSE + a), where Fitness is the individual fitness value, MSE is the mean square error, and a is a very small constant to avoid division by zero error. The fitness value drives the population to evolve in the direction of minimizing the error.

[0084] Step 6: The genetic evolution process. The population is optimized through three-stage operations: In the first stage, a probability selection mechanism based on geometric distribution is used to preferentially select individuals with high fitness to maintain population diversity. In the second stage, an arithmetic crossover strategy is adopted to generate new individuals, and their offspring are generated by the linear combination of the parents, aiming to enhance the exploration ability of the solution space. In the third stage, a non-uniform mutation strategy is introduced, and the mutation intensity decays exponentially with the increase in the number of iterations to ensure the stability of the search in the later stage of the algorithm. This design balances global exploration and local development and effectively improves the convergence efficiency.

[0085] Step 7: Determine the termination condition. If the genetic algorithm reaches the termination condition, the genetic algorithm is terminated and the optimal individual (i.e., the optimal chromosome) is output. The termination condition is, for example, the maximum number of generations (e.g., 50 generations) or the convergence of the fitness value (e.g., the change rate is less than the threshold). If the termination condition is not reached, return to Step 6 to enter the genetic evolution process until the termination condition is reached, and then output the optimal individual (i.e., the optimal chromosome).

[0086] Correspondingly, the processing process of the backpropagation neural network includes:

[0087] Step 1: Decode the optimal individual, and update the network structure parameters based on the optimal connection weights between the input layer and the hidden layer, the optimal critical activation thresholds of the neurons in the hidden layer, the optimal connection weights between the hidden layer and the output layer, and the optimal critical activation thresholds of the neurons in the output layer obtained by decoding.

[0088] Decode the optimal individual into the weights and thresholds (which can also be understood as biases) of the backpropagation neural network, and assign them to the backpropagation neural network to initialize the parameters of the backpropagation neural network.

[0089] Then, train the backpropagation neural network. Specifically, adopt the backpropagation algorithm, set the maximum number of iterations to 1000 times for example, the target error to 1e-6 for example, and the learning rate to 0.01 for example. Update the weights and thresholds through gradient descent; if the training error reaches the target error (1e-6) or reaches the maximum number of iterations, stop training. Otherwise, continue to update the weights and biases.

[0090] In the training method of the above distribution network carrying capacity evaluation model, a genetic algorithm is introduced on the basis of the backpropagation neural network to construct an initial distribution network carrying capacity evaluation model, namely the GA-BP model, aiming to overcome the limitation that the BP neural network is prone to fall into local optimal solutions. Through the intervention of the genetic algorithm, the randomness of the assignment of the weights and thresholds of the backpropagation neural network is effectively reduced, and the training efficiency of the model is significantly improved. The backpropagation neural network optimized by the genetic algorithm not only enhances the adaptability and convergence of the original backpropagation neural network, but also greatly improves the recognition accuracy of the backpropagation neural network, so that the trained distribution network carrying capacity evaluation model exhibits more excellent performance characteristics.

[0091] In some embodiments, the process of training the initial distribution network carrying capacity evaluation model is further described.

[0092] When the preprocessing network is constructed based on the sparrow search algorithm, the fitness function of the preprocessing network is used to take the reciprocal of the prediction error of the backpropagation neural network as the individual fitness value.

[0093] Meanwhile, please refer to Figure 4 The flowchart of the sparrow search algorithm for optimizing the backpropagation neural network shown. In the process of training the initial distribution network carrying capacity evaluation model, the processing process of the preprocessing network includes:

[0094] Step 1: Taking the number obtained by adding the number of input layer nodes, the number of hidden layer nodes and the number of output layer nodes in the back propagation neural network as the number of sparrow individuals in the initial population, and constructing the initial population.

[0095] The first step is initialization. At the beginning of the sparrow search algorithm, the input data is normalized. Through a specific normalization function, the convergence efficiency and stability of the subsequent algorithm are improved. At the same time, the population position of 20 sparrow individuals is initialized.

[0096] Step 2: Encode the population position of the individual sparrows in the population into a vector. The vector corresponds to the connection weight between the input layer and the hidden layer in the back-propagation neural network, the critical activation threshold of the neurons in the hidden layer, the connection weight between the hidden layer and the output layer, and the critical activation threshold of the neurons in the output layer.

[0097] The position of each individual sparrow is encoded as a 205-dimensional vector, which accurately corresponds to the weights and biases of the back-propagation neural network. The specific calculation formula is 14×10+10×5+10+5, where 14 is the number of input layer nodes (corresponding to the feature dimension), 10 is the number of hidden layer nodes, and 5 is the number of output layer nodes (corresponding to the number of categories).

[0098] Step three, determine the individual fitness of the sparrow individuals in the initial population based on the fitness function, and iteratively optimize the initial population based on the individual fitness of each individual until the termination condition is reached, and output the optimal vector of the sparrow individuals in the optimized population.

[0099] The second step is the iterative process of the sparrow search algorithm. In the process of iterative optimization of the back propagation neural network by the sparrow search algorithm, the discoverer relies on the global search capability to explore new areas and update the position according to the discoverer-follower mechanism, guiding the population to move towards the potential better solution area. Based on the discoverer's position, the follower deeply explores the better area to complete its own position update, realizing local search and development. The vigilant balances the exploration and development of the algorithm to prevent the population from falling into the local optimum too early. At the same time, the inverse of the prediction error of the back propagation neural network is used to evaluate the individual fitness, find and update the optimal sparrow position, and provide a high-quality starting point for subsequent iterations.

[0100] The third step is to determine the termination condition. Determine whether the preset maximum number of iterations (set to 50 in this embodiment) has been reached or other termination conditions have been met (such as the fitness value change in multiple consecutive iterations is less than a preset threshold, etc.). If the termination condition is met, proceed to the next step. Otherwise, return to continue the iterative process of the sparrow search algorithm until the termination condition is met.

[0101] The fourth step is to determine the optimal weight and threshold of the back propagation neural network. When the termination condition is met, the weight and bias corresponding to the optimal sparrow position are assigned to the back propagation neural network as its optimal weight and threshold. This operation ensures that the back propagation neural network can be carried out based on the optimized parameters in the subsequent training and prediction process, which is expected to improve the performance of the network.

[0102] Correspondingly, the processing process of the back propagation neural network includes:

[0103] Step 1: decode the optimal vector, and update the network structure parameters based on the decoded optimal connection weights between the input layer and the hidden layer, the optimal critical activation threshold of the neurons in the hidden layer, the optimal connection weights between the hidden layer and the output layer, and the optimal critical activation threshold of the neurons in the output layer.

[0104] The fifth step is back propagation neural network training. With the determined optimal weights and thresholds, combined with the pre-set network structure (14 nodes in the input layer, 10 nodes in the hidden layer, and 5 nodes in the output layer), as well as the hyperbolic tangent function (tansig) used in the hidden layer and the linear function (purelin) used in the output layer, the back propagation neural network is trained. During the training process, the network continuously adjusts the weights and biases through the back propagation algorithm to minimize the prediction error.

[0105] Step 6: Simulation prediction. Use the trained back propagation neural network to perform simulation prediction, and output the prediction results corresponding to 5 categories for the input data. By analyzing and evaluating the prediction results, the performance of the back propagation neural network optimized by SSA can be tested in practical applications.

[0106] In the training method of the distribution network carrying capacity assessment model mentioned above, the sparrow search algorithm is introduced on the basis of the back propagation neural network, and the initial distribution network carrying capacity assessment model is constructed as the SSA-BP model. When optimizing the back propagation neural network, the sparrow search algorithm optimizes the weights and biases of the back propagation neural network by simulating the foraging and anti-predation behavior of the sparrow population. In the algorithm, the sparrow population is divided into producers and foragers. The producers are responsible for global search and the foragers are responsible for local search. The two work together to balance exploration and utilization. When the natural enemy appears, the sparrow population will adjust its position through the alarm signal to avoid falling into the local optimum. In this way, the sparrow search algorithm dynamically adjusts the parameters of the back propagation neural network in the solution space, thereby improving the convergence speed and prediction accuracy of the network, and can effectively avoid the problem of the back propagation neural network falling into the local optimal solution.

[0107] In some embodiments, the backpropagation neural network includes an input layer, an output layer, and a hidden layer. Among them, the hidden layer includes 3 neurons, the activation function of the hidden layer is the hyperbolic tangent function, and the activation function of the output layer is a linear function.

[0108] In some embodiments, see Figure 5 , step 206 trains the initial distribution network carrying capacity evaluation model based on the data of the distribution network under each carrying capacity evaluation index, including:

[0109] Step 502 combines the data of the distribution network under each carrying capacity evaluation index into a data set, and divides the data set into a training set and a test set.

[0110] 80% of the data in the data set can be divided into the training set, and the remaining 20% is divided into the test set.

[0111] Step 504 uses the data in the test set as the initial input data, normalizes the initial input data to the data within the range of 0 to 1, and adds random noise to the normalized data to obtain the input data.

[0112] The initial input data is normalized to the [0, 1] interval by Min-Max normalization to eliminate the dimension difference.

[0113] To increase the diversity and authenticity of the data, random noise is added to each data based on the initial input data. The range of the noise can be between -10 to 10 and -20 to 20 respectively. Although the noise is also randomly generated, the value of the noise is relatively fixed, reflecting the unique attributes or labels of certain evaluation indexes.

[0114] Step 506 trains the initial distribution network carrying capacity evaluation model according to the input data and the test set.

[0115] The training result of the initial distribution network carrying capacity evaluation model can be determined according to the input data and the test data, and then the initial distribution network carrying capacity evaluation model can be continuously optimized according to the determined training result until the iteration stop condition is reached, and then the distribution network carrying capacity evaluation model is obtained.

[0116] The data set generated by the method provided in this embodiment not only has randomness and diversity, but also simulates the uncertainty in the actual data by adding noise, thereby helping to improve the robustness and generalization ability of the distribution network carrying capacity evaluation model.

[0117] In some embodiments, a subjective weighting method or an objective weighting method can be adopted to combine the distribution network carrying capacity evaluation model constructed based on the genetic algorithm and the distribution network carrying capacity evaluation model constructed based on the sparrow search algorithm. Through the combined combined model, the best evaluation result of the distribution network carrying capacity can be obtained.

[0118] Adopt the subjective weighting method. For example, according to the performance evaluation of the genetic algorithm and the sparrow search algorithm by experts in similar-level evaluation problems, an initial weight assignment is given. The objective weighting method can be used to determine the weights according to the evaluation indexes of the two models on the test set. For example, calculate the proportion of the accuracy of the two models in the total accuracy (the sum of the accuracies of the two models) as the preliminary objective weight, and finally use the game theory combined weighting to obtain the subjective and objective combined weight value.

[0119] For example, let the distribution network carrying capacity evaluation model constructed based on the genetic algorithm be M1, and the distribution network carrying capacity evaluation model constructed based on the sparrow search algorithm be M2. The weight of M1 obtained by the combined weighting method is w1, and the weight of M2 is w2. Among them, w1 + w2 = 1. For a sample x to be evaluated, the output y of the combined model can be obtained through the following formula: y = w1 * M(x) + w2 * M(x), so that the models optimized by the two algorithms are combined for level evaluation. In practical applications, it is necessary to adjust and optimize the weights of the combined weighting method according to the specific evaluation results and sample situations to achieve the best level evaluation effect.

[0120] Please refer to Figure 6 , some embodiments of the present application also provide a method for evaluating the carrying capacity of a distribution network, including:

[0121] Step 602, obtain the distribution network carrying capacity evaluation model trained according to the above-mentioned training method of the distribution network carrying capacity evaluation model.

[0122] For the description of the training method of the distribution network carrying capacity evaluation model, reference can be made to the above relevant description, which will not be elaborated here.

[0123] In the training method of the above-mentioned distribution network carrying capacity evaluation model, a genetic algorithm is introduced on the basis of the backpropagation neural network to construct an initial distribution network carrying capacity evaluation model, namely the GA-BP model, aiming to overcome the limitation that the BP neural network is prone to falling into local optimal solutions. Through the intervention of the genetic algorithm, the randomness of the weight and threshold assignment of the backpropagation neural network is effectively reduced, and the training efficiency of the model is significantly improved. After being optimized by the genetic algorithm, the backpropagation neural network not only enhances the adaptability and convergence of the original backpropagation neural network, but also greatly improves the recognition accuracy of the backpropagation neural network, so that the trained distribution network carrying capacity evaluation model exhibits more excellent performance characteristics.

[0124] On the basis of the backpropagation neural network, a sparrow search algorithm is introduced to construct an initial distribution network carrying capacity evaluation model, namely the SSA-BP model. When optimizing the backpropagation neural network, the sparrow search algorithm optimizes the weights and biases of the backpropagation neural network by simulating the foraging and anti-predation behaviors of the sparrow population. In the algorithm, the sparrow population is divided into producers and foragers. The producers are responsible for global search, and the foragers are responsible for local search. The two work together to balance exploration and exploitation. When a natural enemy appears, the sparrow population will adjust its position through an alarm signal to avoid falling into local optimality. In this way, the sparrow search algorithm dynamically adjusts the parameters of the backpropagation neural network in the solution space, thereby improving the convergence speed and prediction accuracy of the network, and effectively avoiding the problem that the backpropagation neural network falls into local optimal solutions.

[0125] Step 604: Obtain the data of the distribution network under each carrying capacity evaluation index after multiple charging piles are incorporated into the distribution network.

[0126] Among them, the carrying capacity evaluation index includes at least one of the passing rate of medium-voltage line N-1, average power outage frequency, average outage duration, evaluation power supply reliability rate, voltage qualification rate, voltage deviation, total harmonic distortion rate, line loss rate, line maximum load rate, line annual average load rate, distribution transformer annual average load rate, net load maximum allowable volatility, net load volatility or load transfer ratio under different carrying capacity levels.

[0127] Step 606: Input the data of the distribution network under each carrying capacity evaluation index into the distribution network carrying capacity evaluation model to obtain the carrying capacity level output by the distribution network carrying capacity evaluation model; among them, the carrying capacity level is positively correlated with the carrying capacity of the distribution network.

[0128] The result output by the distribution network carrying capacity evaluation model is the carrying capacity level of the distribution network. The larger the carrying capacity level, the stronger the carrying capacity of the distribution network. The smaller the carrying capacity level, the weaker the carrying capacity of the distribution network. Therefore, by using the method provided in this embodiment, the impact of the supercharger on the distribution network after grid connection can be effectively evaluated in a hierarchical and quantitative manner, which is convenient for subsequent reference in the layout of superchargers.

[0129] The method provided in this embodiment solves some limitations of the backpropagation neural network in practical applications. If the initial weights and thresholds are randomly set, it is easy to fall into a local optimal solution, resulting in limited performance of the distribution network carrying capacity evaluation model. The performance of the distribution network carrying capacity evaluation model is improved, making the results obtained when using the distribution network carrying capacity evaluation model to evaluate the carrying capacity of the distribution network more objective and accurate. Compared with the inaccurate evaluation of the distribution network carrying capacity in the traditional scheme, the method provided in this embodiment improves the accuracy of the evaluation of the distribution network carrying capacity.

[0130] It should be understood that although the steps in the flowcharts involved in the above embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this embodiment, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or steps in other steps.

[0131] Based on the same inventive concept, the embodiment of the present application also provides a training device for a distribution network carrying capacity evaluation model for implementing the training method of the distribution network carrying capacity evaluation model involved above. The implementation solution for solving problems provided by this device is similar to the implementation solution described in the above method. Therefore, the specific limitations in one or more embodiments of the training device for the distribution network carrying capacity evaluation model provided below can refer to the limitations on the training method of the distribution network carrying capacity evaluation model in the above text, and will not be repeated here.

[0132] In an exemplary embodiment, as Figure 7 shown, a training device 70 for a distribution network carrying capacity evaluation model is provided, including: a model construction module 71, an acquisition module 72, and a training module 73, where:

[0133] A model construction module 71 for constructing an initial evaluation model for the carrying capacity of a distribution network. The initial evaluation model for the carrying capacity of a distribution network includes a pre-processing network and a backpropagation neural network, and the output of the pre-processing network is used as the input of the backpropagation neural network. The pre-processing network is constructed based on a genetic algorithm or a sparrow search algorithm.

[0134] An acquisition module 72 for acquiring data of the distribution network under each carrying capacity evaluation index after multiple charging piles are incorporated into the distribution network. Among them, the carrying capacity evaluation index includes at least one of the passing rate of medium-voltage line N-1, average power outage frequency, average outage duration, evaluated power supply reliability rate, voltage qualification rate, voltage deviation, total harmonic distortion rate, line loss rate, maximum line load rate, annual average line load rate, annual average distribution transformer load rate, maximum allowable volatility of net load, net load volatility, or load transfer ratio under different carrying capacity levels.

[0135] A training module 73 for training the initial evaluation model for the carrying capacity of a distribution network based on the data of the distribution network under each carrying capacity evaluation index, and obtaining an evaluation model for the carrying capacity of a distribution network after the training is completed.

[0136] In one embodiment, when the pre-processing network is constructed based on a genetic algorithm, the fitness function of the pre-processing network is used to take the reciprocal of the prediction error of the backpropagation neural network as the individual fitness value.

[0137] During the process of training the initial evaluation model for the carrying capacity of a distribution network, the processing process of the pre-processing network includes:

[0138] According to the coding range of [-1, 1], the connection weights between the input layer and the hidden layer, the critical activation thresholds of the neurons in the hidden layer, the connection weights between the hidden layer and the output layer, and the critical activation thresholds of the neurons in the output layer in the backpropagation neural network are respectively encoded into chromosomes, and an initial population is formed based on the encoded chromosomes. Among them, an individual in the population corresponds to a chromosome.

[0139] Determine the individual fitness of each individual in the initial population based on the fitness function, and iteratively optimize the initial population based on the individual fitness of each individual until the termination condition is reached, and then output the optimal individual in the optimized population.

[0140] Correspondingly, the processing process of the backpropagation neural network includes:

[0141] Decode the optimal individual, and update the network structure parameters based on the optimal connection weights between the input layer and the hidden layer, the optimal critical activation thresholds of the neurons in the hidden layer, the optimal connection weights between the hidden layer and the output layer, and the optimal critical activation thresholds of the neurons in the output layer obtained by decoding.

[0142] In one embodiment, when the preprocessing network is constructed based on the sparrow search algorithm, the fitness function of the preprocessing network is used to take the reciprocal of the prediction error of the backpropagation neural network as the individual fitness value;

[0143] During the process of training the initial distribution network carrying capacity evaluation model, the processing process of the preprocessing network includes:

[0144] Taking the sum of the number of nodes in the input layer, hidden layer, and output layer in the backpropagation neural network as the number of sparrow individuals in the initial population, and constructing the initial population;

[0145] Encoding the population positions of the sparrow individuals in the population as vectors; the vectors correspond to the connection weights between the input layer and the hidden layer, the critical activation thresholds of the neurons in the hidden layer, the connection weights between the hidden layer and the output layer, and the critical activation thresholds of the neurons in the output layer in the backpropagation neural network;

[0146] Determining the individual fitness of the sparrow individuals in the initial population based on the fitness function, and iteratively optimizing the initial population based on the individual fitness of each individual until the termination condition is reached, and then outputting the optimal vectors of the sparrow individuals in the optimized population;

[0147] Correspondingly, the processing process of the backpropagation neural network includes:

[0148] Decoding the optimal vector, and updating the network structure parameters based on the optimal connection weights between the input layer and the hidden layer, the optimal critical activation thresholds of the neurons in the hidden layer, the optimal connection weights between the hidden layer and the output layer, and the optimal critical activation thresholds of the neurons in the output layer obtained by decoding.

[0149] In one embodiment, the backpropagation neural network includes an input layer, an output layer, and one hidden layer; wherein, the hidden layer includes 3 neurons, the activation function of the hidden layer is the hyperbolic tangent function, and the activation function of the output layer is a linear function.

[0150] In one embodiment, the training module 73 is used to merge the data of the distribution network under each carrying capacity evaluation index into a data set, and divide the data set into a training set and a test set; use the data in the test set as the initial input data, and normalize the initial input data to the data within the range composed of 0 to 1, and add random noise to the normalized data to obtain the input data; train the initial distribution network carrying capacity evaluation model according to the input data and the test set.

[0151] Each module in the training device of the above-mentioned distribution network carrying capacity evaluation model can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0152] In an exemplary embodiment, as Figure 8 shown, a distribution network carrying capacity evaluation device 80 is provided, including an acquisition module 81 and a processing module 82, where:

[0153] The acquisition module 81 acquires the distribution network carrying capacity evaluation model trained according to the above-mentioned training method of the distribution network carrying capacity evaluation model.

[0154] The acquisition module 81 is further configured to acquire data of the distribution network under each carrying capacity evaluation index after multiple charging piles are incorporated into the distribution network; wherein, the carrying capacity evaluation index includes at least one of the passing rate of medium-voltage line N-1, average power outage frequency, average outage duration, evaluation power supply reliability rate, voltage qualification rate, voltage deviation, total harmonic distortion rate, line loss rate, line maximum load rate, line annual average load rate, distribution transformer annual average load rate, net load maximum allowable volatility, net load volatility, or load transfer ratio under different carrying capacity levels.

[0155] The processing module 82 inputs the data of the distribution network under each carrying capacity evaluation index into the distribution network carrying capacity evaluation model to obtain the carrying capacity level output by the distribution network carrying capacity evaluation model; wherein, the carrying capacity level is positively correlated with the distribution network carrying capacity.

[0156] Each module in the above-mentioned distribution network carrying capacity evaluation device can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0157] In an exemplary embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 9As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data of the distribution network under each carrying capacity assessment. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it implements the training method of the distribution network carrying capacity assessment model provided in any one of the above embodiments, or implements the distribution network carrying capacity assessment method provided in any one of the above embodiments.

[0158] Those skilled in the art can understand that Figure 9 the structure shown in is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0159] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, it implements the training method of the distribution network carrying capacity assessment model provided in any one of the above embodiments, or implements the distribution network carrying capacity assessment method provided in any one of the above embodiments.

[0160] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, it implements the training method of the distribution network carrying capacity assessment model provided in any one of the above embodiments, or implements the distribution network carrying capacity assessment method provided in any one of the above embodiments.

[0161] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. This computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, artificial intelligence (AI) processors, etc., without limitation.

[0162] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in the present application.

[0163] The above embodiments only represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A training method for a distribution network carrying capacity assessment model, characterized in that: The method comprises: Constructing an initial distribution network carrying capacity assessment model, wherein the initial distribution network carrying capacity assessment model includes a pre-order processing network and a back-propagation neural network, wherein the output of the pre-order processing network is used as the input of the back-propagation neural network; the pre-order processing network is constructed based on a genetic algorithm or a sparrow search algorithm; After multiple charging piles are incorporated into the distribution network, the data of the distribution network under each carrying capacity evaluation index is obtained; wherein the carrying capacity evaluation index includes at least one of the medium voltage line N-1 pass rate, average power outage frequency, average outage duration, evaluated power supply reliability rate, voltage qualification rate, voltage deviation, total harmonic distortion rate, line loss rate, line maximum load rate, line annual average load rate, distribution transformer annual average load rate, net load maximum allowable fluctuation rate, net load fluctuation rate or load transfer ratio under different carrying capacity levels; Based on the data of the distribution network under each carrying capacity evaluation index, the initial distribution network carrying capacity evaluation model is trained, and the distribution network carrying capacity evaluation model is obtained after the training is completed.

2. The method according to claim 1, characterized in that In the case where the pre-order processing network is constructed based on a genetic algorithm, the fitness function of the pre-order processing network is used to use the inverse of the prediction error of the back propagation neural network as the individual fitness value; In the process of training the initial distribution network carrying capacity assessment model, the processing process of the pre-processing network includes: According to the coding range of [-1,1], the connection weights between the input layer and the hidden layer, the critical activation threshold of the neurons in the hidden layer, the connection weights between the hidden layer and the output layer, and the critical activation threshold of the neurons in the output layer in the back propagation neural network are respectively encoded as chromosomes, and an initial population is formed based on the encoded chromosomes; wherein one individual in the population corresponds to one chromosome; Determine the individual fitness of each individual in the initial population based on the fitness function, and iteratively optimize the initial population based on the individual fitness of each individual until the termination condition is reached, and output the optimal individual in the optimized population; Correspondingly, the processing process of the back propagation neural network includes: Decode the optimal individual, and update the network structure parameters based on the decoded optimal connection weights between the input layer and the hidden layer, the optimal critical activation threshold of the neurons in the hidden layer, the optimal connection weights between the hidden layer and the output layer, and the optimal critical activation threshold of the neurons in the output layer.

3. The method according to claim 1, characterized in that: In the case where the pre-order processing network is constructed based on a sparrow search algorithm, the fitness function of the pre-order processing network is used to use the inverse of the prediction error of the back propagation neural network as the individual fitness value; In the process of training the initial distribution network carrying capacity assessment model, the processing process of the pre-processing network includes: The number obtained by adding the number of input layer nodes, the number of hidden layer nodes and the number of output layer nodes in the back propagation neural network is used as the number of sparrow individuals in the initial population, and the initial population is constructed; Encoding the population position of the individual sparrows in the population into a vector; the vector corresponds to the connection weight between the input layer and the hidden layer, the critical activation threshold of the neurons in the hidden layer, the connection weight between the hidden layer and the output layer, and the critical activation threshold of the neurons in the output layer in the back propagation neural network; The individual fitness of the sparrow individuals in the initial population is determined based on the fitness function, and the initial population is iteratively optimized based on the individual fitness of each individual until the termination condition is reached, and the optimal vector of the sparrow individuals in the optimized population is output; Correspondingly, the processing process of the back propagation neural network includes: The optimal vector is decoded, and the network structure parameters are updated based on the decoded optimal connection weights between the input layer and the hidden layer, the optimal critical activation threshold of the neurons in the hidden layer, the optimal connection weights between the hidden layer and the output layer, and the optimal critical activation threshold of the neurons in the output layer.

4. The method according to any one of claims 1 to 3, characterized in that: The back propagation neural network includes an input layer, an output layer and a hidden layer; wherein the hidden layer includes 3 neurons, the activation function of the hidden layer is a hyperbolic tangent function, and the activation function of the output layer is a linear function.

5. The method according to any one of claims 1 to 3, characterized in that: The training of the initial distribution network carrying capacity evaluation model based on the data of the distribution network under each carrying capacity evaluation index includes: The data of the distribution network under each carrying capacity evaluation indicator are merged into a data set, and the data set is divided into a training set and a test set; The data in the test set is used as the initial input data, and the initial input data is normalized into data within the interval range of 0 to 1, and random noise is added to the normalized data to obtain the input data; The initial distribution network carrying capacity assessment model is trained according to the input data and the test set.

6. A method for evaluating the carrying capacity of a distribution network, characterized in that: The method comprises: Obtaining a distribution network carrying capacity assessment model training method according to any one of claims 1 to 5, and a distribution network carrying capacity assessment model obtained by training; After multiple charging piles are incorporated into the distribution network, the data of the distribution network under each carrying capacity evaluation index is obtained; wherein the carrying capacity evaluation index includes at least one of the medium voltage line N-1 pass rate, average power outage frequency, average outage duration, evaluated power supply reliability rate, voltage qualification rate, voltage deviation, total harmonic distortion rate, line loss rate, line maximum load rate, line annual average load rate, distribution transformer annual average load rate, net load maximum allowable fluctuation rate, net load fluctuation rate or load transfer ratio under different carrying capacity levels; The data of the distribution network under each carrying capacity assessment indicator is input into the distribution network carrying capacity assessment model to obtain the carrying capacity level output by the distribution network carrying capacity assessment model; wherein the carrying capacity level is positively correlated with the distribution network carrying capacity.

7. A training device for a distribution network carrying capacity assessment model, characterized in that: The device comprises: A model building module, used to build an initial distribution network carrying capacity assessment model, wherein the initial distribution network carrying capacity assessment model includes a pre-order processing network and a back-propagation neural network, wherein the output of the pre-order processing network is used as the input of the back-propagation neural network; the pre-order processing network is built based on a genetic algorithm or a sparrow search algorithm; An acquisition module is used to obtain data of the distribution network under each carrying capacity evaluation index after multiple charging piles are incorporated into the distribution network; wherein the carrying capacity evaluation index includes at least one of the medium voltage line N-1 pass rate, average power outage frequency, average outage duration, evaluated power supply reliability rate, voltage qualification rate, voltage deviation, total harmonic distortion rate, line loss rate, line maximum load rate, line annual average load rate, distribution transformer annual average load rate, net load maximum allowable fluctuation rate, net load fluctuation rate or load transfer ratio under different carrying capacity levels; The training module is used to train the initial distribution network carrying capacity evaluation model based on the data of the distribution network under each carrying capacity evaluation indicator, and obtain the distribution network carrying capacity evaluation model after the training is completed.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the processor implements the steps of the method according to any one of claims 1 to 5, or implements the steps of the method according to claim 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 5, or implements the steps of the method according to claim 6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 5, or implements the steps of the method according to claim 6.

Images